Uncertainty in Catchment Delineations as a Result of Digital Elevation Model Choice

Nine digital elevation model (DEM) datasets were used for separate delineations of the Nam Co, Tibet catchment and its subcatchments, and these delineated areas were compared using the highest resolution dataset, TanDEM-X 12 m, as a baseline. The mean delineated catchment area was within 0.1% percent of the baseline delineation, with a standard error of the mean (SEM) that was 0.13% of the baseline. In a comparison of 49 subcatchment areas, TanDEM-X and ALOS datasets delineated similar areas, followed closely by SRTM 30 m, then SRTM 90 m, ACE2, and ASTER GDEM1. ASTER GDEM2 was a noteworthy outlier, having the largest mean subcatchment area that was nearly three times that of the baseline mean. Correlation coefficients were calculated for subcatchment parameters, SEM, and each DEM’s subcatchment area error. SEM had a weak but significant negative correlation with the mean and median slope. ASTER GDEM1 and GDEM2 were the only datasets that showed any significant correlations with the subcatchment environment variables, though these correlations were also weak. The 30 m posting ASTER GDEMs performed worse against the baseline than the other 30 m and 90 m datasets, showing that posting alone does not determine how good a dataset is. Our results show general small errors for catchment delineations, though there is the possibility for large errors, particularly in the older ASTER and SRTM datasets.